[ARCHIVE] Availability metrics and latency

Archived 2022.11.20. Content moved to error scenarios and responses.

Question

The CDS Availability Requirements NFR provides the statement:

“The definition of a period of unavailability is any period of time when any of the API end points defined in the standard is unable to reliably provide a successful response to an appropriately constructed request.”

Is it  necessary to consider data latency in determining whether an API end point is providing a "successful response"?

Data latency in the above context means a difference between the values returned in a CDR API response and what our source systems may hold for the equivalent data at that point in time.

As an example for context; if our API was ‘reliably’ responding to requests, but returning cached data from 2 hours ago, would that be considered as "available" and "successfully responding". There are two cases to consider:

• Other primary digital channels also reflected the same delay (i.e. the API response is the same as other channels)
• Other primary channels did not have the same delay (i.e. the API response is delayed compared to other channels)

Answer

Latency that affects the quality of data may not be considered unavailability, but it would still be in breach of the data latency requirements in the NFR and should not be considered a successful outcome.